Hi @vchung ,
My submission seems to have gone through and been evaluated "successfully". But the QWK for ADNC, Braak and Thal are all exactly 0 and there are no entries for their R2 values. Can you kindly verify that it was evaluated correctly, in case you have access to the predictions.csv file?
(I only see some STDERR in the log file at the very beginning of the run, and I'm not sure if it has anything to do with the outcome I'm seeing.)
Can you also please clarify if the Donor IDs in data.h5ad file are encoded as utf-8 like in the training data? I'm wondering if there could be an issue in the Donor IDs I write to predictions.csv.
Thank you!
-Nikhil
Created by Nikhil Karthik nkck Hi @nkck,
Yes, all of these are available / OK to use!
Best,
Kyle Thank you for the clarification! @ktravaglini Can you kindly confirm if the following entries are available in data.h5ad?
- obs
- APOE Genotype
- Age at Death
- Hispanic
- Race (choice=American Indian)
- Race (choice=Asian)
- Race (choice=Black)
- Race (choice=Native Hawaiian or Pacific Islander)
- Race (choice=Other)
- Race (choice=Unknown or unreported)
- Race (choice=White)
- Sex
I saw in another post that Cognitive Status is unavailable. So I just wanted to make sure that these are fine.
Thank you so much!
Nikhil Hi @nkck ,
I forwarded your question to one of the challenge organizers and they have confirmed that the ordering is the same.
I am tagging @ktravaglini , in case you have any additional follow-up questions regarding the data. Thanks! Hi @vchung,
Thanks a lot again for your help and time. Would it be possible to say whether the ordering of the 36601 genes ['MIR1302-2HG', 'FAM138A', 'OR4F5', 'AL627309.1', 'AL627309.3', 'AL627309.2', 'AL627309.5', ....] are exactly the same in the training and the test datasets? To be specific, if I convert var/gene_ids to a list, would 'MIR1302-2HG' be at index 0, etc exactly as in training dataset? Since the ordering was fixed in both A9 and MTG datasets, I thought this was true.
Sorry for the additional query, but it will help me understand whether there is a mismatch between training and test datasets due to my implicit assumption.
Thanks!
Nikhil Oh I see! Thank you @vchung! Hi @nkck ,
Thanks for reaching out! We do have access to your generated predictions file. From what I can see, almost all of the predictions generated by subID 9760279 are the same value per column, thus explaining why the QWK and R^2 are 0.
Hope this helps!
EDIT: add submission ID